Basic example

Author

Jono Tuke

Published

April 1, 2023

This document is based on the r-code from Jacinta. It is for me to understand how the code works, and also to verify with Jacinta that my understanding is correct.

The file from Jacinta is Simulate clouds.R and is in inst/notes.

Read in raster

We read in the raster using the raster function from the raster package.

pacman::p_load(tidyverse, raster)
example_raster <- raster(
  here::here("inst", "notes", "InjuneCloudFree27Sept.tif")
)
example_raster

class      : RasterLayer 
dimensions : 7061, 7911, 55859571  (nrow, ncol, ncell)
resolution : 30, 30  (x, y)
extent     : 426885, 664215, 7019185, 7231015  (xmin, xmax, ymin, ymax)
crs        : +proj=utm +zone=55 +south +ellps=WGS84 +units=m +no_defs 
source     : InjuneCloudFree27Sept.tif 
names      : InjuneCloudFree27Sept

The basic form of a raster is a grid of values with associated spatial information like lat/long and projection. Note that this is an S4 object. The information of importance is

dimensions: size of the grid
resolution: geographical size of each grid, in this case I assume 30 metres by 30 metres.
extent: spatial position.
crs: projection system.

We can plot it with the plot function.

plot(example_raster)

Simulate clouds

To simulate the clouds, we first set up an appropriate sized grid.

cloud_size <- 70
cloud_grid <- expand_grid(
  X = 1:cloud_size, 
  Y = 1:cloud_size
)
cloud_grid

# A tibble: 4,900 × 2
       X     Y
   <int> <int>
 1     1     1
 2     1     2
 3     1     3
 4     1     4
 5     1     5
 6     1     6
 7     1     7
 8     1     8
 9     1     9
10     1    10
# ℹ 4,890 more rows

We will get the number of cells

grid_size <- nrow(cloud_grid)

So we have 4900 cells in our cloud_grid.

Next, we will get the distances between each point

distance <- as.matrix(dist(cloud_grid))
distance[1:5, 1:5]

  1 2 3 4 5
1 0 1 2 3 4
2 1 0 1 2 3
3 2 1 0 1 2
4 3 2 1 0 1
5 4 3 2 1 0

image(distance[1:10, 1:10])

The simulation of the cloud is done with this code

phi <- 0.1
set.seed(2023)
cloud_grid$Z <- mgcv::rmvn(1, rep(0, grid_size), exp(-phi * distance))

So we have a single random variate from a normal distribution with the distribution

\[ X \sim N_n(\boldsymbol{0}, \Sigma), \]

where

\[ [\Sigma]_{ij} = \exp(-\phi d_{ij}), \] and \(d_{ij}\) is the Euclidean distance between \(X_i\) and \(X_j\), and we have \(n\) points.

We can have a gander:

cloud_grid |> 
  ggplot(aes(X, Y, fill = Z)) + 
  geom_tile()

Next, we convert to a raster

cloud_raster <- rasterFromXYZ(cloud_grid)
plot(cloud_raster)

plot(example_raster)

Matching cloud raster to original raster

So that the cloud raster and example raster are at the same location, we need to set the cloud raster extent and crs to be the same.

cloud_raster

class      : RasterLayer 
dimensions : 70, 70, 4900  (nrow, ncol, ncell)
resolution : 1, 1  (x, y)
extent     : 0.5, 70.5, 0.5, 70.5  (xmin, xmax, ymin, ymax)
crs        : NA 
source     : memory
names      : Z 
values     : -2.595866, 2.997786  (min, max)

extent(cloud_raster) <- extent(example_raster)
crs(cloud_raster) <- proj4string(example_raster)
cloud_raster

class      : RasterLayer 
dimensions : 70, 70, 4900  (nrow, ncol, ncell)
resolution : 3390.429, 3026.143  (x, y)
extent     : 426885, 664215, 7019185, 7231015  (xmin, xmax, ymin, ymax)
crs        : +proj=utm +zone=55 +south +ellps=WGS84 +units=m +no_defs 
source     : memory
names      : Z 
values     : -2.595866, 2.997786  (min, max)

plot(cloud_raster)

Next, we notice that the dimension of the example and cloud are not the same, so we match them using projectRaster()

cloud_raster <- projectRaster(cloud_raster, example_raster, method = 'ngb')
cloud_raster

class      : RasterLayer 
dimensions : 7061, 7911, 55859571  (nrow, ncol, ncell)
resolution : 30, 30  (x, y)
extent     : 426885, 664215, 7019185, 7231015  (xmin, xmax, ymin, ymax)
crs        : +proj=utm +zone=55 +south +ellps=WGS84 +units=m +no_defs 
source     : memory
names      : Z 
values     : -2.595866, 2.997786  (min, max)

plot(cloud_raster)

Filtering cloud

We can use different ranges to set to zero to change the cloud formation:

speckled small clouds

speckled_cloud_raster <- cloud_raster
speckled_cloud_raster[speckled_cloud_raster > -0.5 & speckled_cloud_raster < 0.15] <- NA
plot(speckled_cloud_raster, colNA = "blue")

Large clouds

large_cloud_raster <- cloud_raster
large_cloud_raster[large_cloud_raster > 0.8 & large_cloud_raster < 4] <- NA
plot(large_cloud_raster, colNA = "blue")

Add clouds to original raster

We can add the clouds to the original raster using mask

example_cloud_raster <- mask(
  example_raster, 
  mask = large_cloud_raster
)
plot(example_raster)

plot(example_cloud_raster)

Writing out

The function writeGDAL is being deprecated:

https://r-spatial.org/r/2022/04/12/evolution.html

So we need to find another way to write out. Maybe this - need to talk to Jacinta about best way.

writeRaster(
  example_cloud_raster, 
  filename = here::here(
    "inst", "notes", "example_cloud_raster.tif"
  ), 
  overwrite = TRUE
)